2 Radar Data Preprocessing
Weather radar data of the firework events at the turns of the years usually contain some degree of precipitation clutter. To filter out precipitation advanced algorithms such as MistNet have been developed, but as we are dealing with dual-pol data here, we can use a simpler and yet robust method using the depolarization ratio (Kilambi, Fabry, and Meunier 2018).
To make sure our processed weather radar data does not contain any significant proportions of precipitation or ground clutter anymore, we process the data as follows:
- We remove electromagnetic interference based on a visual inspection of the scan and throw out all data of affected rays.
- We calculate the depolarization ratio (Kilambi, Fabry, and Meunier 2018) and separate biology from meteorology by classifying all range gates with a depolarization ratio \(>-12dB\) as biology. We subsequently ‘despeckle’ this, to remove obvious misclassifications.
- We average reflectivity over a number of scans before the time of the fireworks event and throw out the range-gates with highest average reflectivities.
All these steps can be undertaken directly on the polar volume data, so we can subsequently directly plug the cleaned up volume into the range-bias correction.
2.1 Setting-up the pre-processing environment
As usual, we use bioRad (Dokter et al. 2019), but this time we include plotly for some interactive plotting.
2.2 Removing electromagnetic interference
We have determined in which scans birds are taking off based on the maximum increase in reflectivity in the scan for each of the involved radars. Let’s now look at these scans to see how much filtering for electromagnetic interference we need to do. The easiest way to determine which rays are subject to this interference is by plotting the scans in polar coordinates (\((r, \alpha)\)), so interference stands out as horizontal lines of more or less constant, or very gradually changing reflectivities. Plotting using plotly makes it easier to identify the specific problematic rays as one can zoom in to identify the exact azimuths \(\alpha\).
The scans we will be using:
- Herwijnen:
RAD_NL62_VOL_NA_201712312310_ODIM.h5 - Den Helder:
RAD_NL61_VOL_NA_201712312305_ODIM.h5
For illustrative purposes we will only illustrate removal of EM interference for the Herwijnen, as the procedure for Den Helder is exactly identical, but this scan contains very little of said clutter.